How to create a confusion matrix and extract the true and false positives and true and false negatives using scikit learn in python ?

Examples of how to create a confusion matrix and infer the true positive, true negative, false positive and false negative values using scikit learn in python ?

Create a confusion matrix with scikit-learn

To create a confusion matrix a solution is to use scikit-learn:

from sklearn.metrics import confusion_matrix

y_true = [1,1,0,0,1]
y_pred = [1,1,1,0,1]

cm = confusion_matrix(y_true, y_pred, labels=[0, 1])

print(cm)

returns

[[1 1]
 [0 3]]

Get tn, fp, fn, tp for a binary classification

tn, fp, fn, tp = confusion_matrix(list(y_true), list(y_pred), labels=[0, 1]).ravel()

print('True Positive', tp)
print('True Negative', tn)
print('False Positive', fp)
print('False Negative', fn)

gives

True Positive 3
True Negative 1
False Positive 1
False Negative 0

Calculate the accuracy score

from sklearn.metrics import accuracy_score

accuracy_score(y_true, y_pred)

gives

0.8

same as doing

acc = (tp+tn) / (tp+tn+fn+fp)

print(acc)

gives

0.8

Calculate the tp, tn, fp and fn rates

tot = cm.sum()

same as

tot = tn+tp+fp+fn

returns here

5

and then

print('True Positive Rate', tp/tot)
print('True Negative Rate', tn/tot)
print('False Positive Rate', fp/tot)
print('False Negative Rate', fn/tot)

returns

True Positive Rate 0.6
True Negative Rate 0.2
False Positive Rate 0.2
False Negative Rate 0.0

Get tn, fp, fn, tp with more than two categories

from sklearn.metrics import confusion_matrix

y_true = [1,1,0,0,1,1,2,2]
y_pred = [1,1,1,0,2,1,2,2]

cm = confusion_matrix(y_true, y_pred, labels=[0, 1])
cm = confusion_matrix(y_true, y_pred)

print(cm)

gives

[[1 1 0]
 [0 3 1]
 [0 0 2]]

and

fp = cm.sum(axis=0) - np.diag(cm)  
fn = cm.sum(axis=1) - np.diag(cm)
tp = np.diag(cm)
tn = cm.sum() - (fp + fn + tp)

print(fp,fn,tp,tn)

returns

[0 1 1] [1 1 0] [1 3 2] [6 3 5]

Get tp, tn,fp, fn for a given category

idx = 0
print(fp[idx], fn[idx], tp[idx], tn[idx])

returns

0 1 1 6

print(fp[idx], fn[idx], tp[idx], tn[idx])

Category 1

idx = 1

returns

1 1 3 3

Category 2

idx = 2

returns

1 0 2 5

Calculate the accuracy score

from sklearn.metrics import accuracy_score

accuracy_score(y_true, y_pred)

gives

0.75

same as

acc = np.diag(cm).sum() / cm.sum()

Plot a confusion matrix with matplotlib and seaborn

import numpy as np
import matplotlib.pyplot as plt
import seaborn as sn
import pandas as pd

import seaborn as sns
import math

from mpl_toolkits.axes_grid1 import make_axes_locatable

import matplotlib as mpl

mpl.style.use('seaborn')

df_cm = pd.DataFrame(cm, 
    index = [i for i in range(cm.shape[0])],
    columns = [i for i in range(cm.shape[1])])

fig = plt.figure()

plt.clf()

ax = fig.add_subplot(111)
ax.set_aspect(1)

cmap = sns.cubehelix_palette(light=1, as_cmap=True)

res = sn.heatmap(df_cm, annot=True, fmt='.2f', cmap=cmap)

res.invert_yaxis()

#plt.yticks([0.5,1.5,2.5], [ '0', '1', '2'],va='center')

plt.title('Confusion Matrix')

plt.savefig('confusion_matrix_1.png', dpi=100, bbox_inches='tight' )

plt.show()

Normalize the confusion matrix

Just add

sum = cm.sum()
cm = cm * 100.0 / ( 1.0 * sum )

and replace

res = sn.heatmap(df_cm, annot=True, fmt='.2f', cmap=cmap)

by

res = sn.heatmap(df_cm, annot=True, vmin=0.0, vmax=100.0, fmt='.2f', cmap=cmap)

References

• Scikit-learn: How to obtain True Positive, True Negative, False Positive and False Negative
• Confusion Matrix